Open AccessAnalysis and Implementation of Enhanced Image Inpainting method using adjustable patch sizes
Publication year - 2021
Publication title -
international journal of emerging trends in engineering research
Language(s) - English
Resource type - Journals
ISSN - 2347-3983
DOI - 10.30534/ijeter/2021/25932021
Subject(s) - inpainting , mean squared error , peak signal to noise ratio , regularization (linguistics) , image (mathematics) , artificial intelligence , mathematics , process (computing) , computer science , computer vision , pattern recognition (psychology) , statistics , operating system
Image inpainting is the process of reconstruction of the damaged image and removal of unwanted objects in an image. In the image inpainting process patch priority andselection of best patch playsa major role. The patch size is also considered for producing good results in the image inpainting. In this paper patch priority is obtained by introducing a regularization factor (ɷ). The best patch selection is acquired by using the Sum of Absolute Difference (SAD) distance method. The results of inpainting are investigated with adjustable patch sizes of 5×5, 7×7, 9×9, 11×11, and 13×13 for the proposed method. The performance of these adjustable patch sizes is observed by using Peak Signal to Noise Ratio (PSNR) and Mean Square Error (MSE). The best suitable patch size for good inpainting is announced based on the values of PSNR and MSE.